Patrick.Duval@inria.fr
In this paper I propose to investigate an new approach to the evolution of html, in order to ease the development of applications based upon content-understanding, such as the combination of (information) services.
A model is sketched where the use of generic markup languages (e.g. sgml-like declarative meta-languages) allows better communications of contents between active entities over the Web (Web client and server applications).
The objective is to ease the development of flexible clients, and the combination of services - through a better readability of contents. This approach should also ease flexible developments of schemes such as: open html extensions, cooperative content presenters, dynamically reconfigurable and loadable clients, cooperative service providers, new metaphors for interactive multimedia interfaces.
Though each of these statements may contain some sort of truth, one most important feature of html and the Web is misunderstood here: the fundamental accent over sharing and providing contents over the net, letting as much as possible the issues of presentation and interaction handling to the domain of content presenters (client browsers). This feature is actually a very fruitful and efficient paradigm for the design of large scale distributed systems.
Contents basically means here:
A Web model in that perspective looks like a large-scale information space (the cyberspace ) populated by content providers and service providers, designable through url's . These servers provide on demand universally understandable contents (i.e. html) to content presenters (end-users browsers) and content&service consumers. These content consumers applications are new types of Web entities exploiting primary material to deliver value-added services.
A this point however, html will not be able to represent all the kinds of contents and metaphors that users will ask for. It is also difficult to imagine that a single language technology (i.e. a standard-language) should be acceptable at this level for every purpose and every environment all over the net.
A real need arises then for the build-up of a simple interdialect (to reuse the metaphor of the internet ), i.e. a dialect able to flexibly integrate new contents, interaction metaphors and language technologies, as they emerge among Web applications.
At this point, an efficient and open solution is to use a generic declarative language, e.g. a generic markup language like SGML. This could be a large class of ``SGML parsable dialects'', based upon some necessary conventions and usage adaptations of SGML, to make it simple and practically usable and interoperable over the Web (see the references section below).
Once many primary (information) services stand on the Web, being able to build high-order by combination services that exploit this primary material in a variety of ways become a natural idea. First examples of such applications based on content-understanding are the WWWWORM and the WebCrawler (information providers that build from other information providers). Providing more generally value added services would be an other example of application of this concept of combination of services.
At this point. improving the dialect(s) used by information services to express contents (information, services interfaces) will be a major requirement. Here again, the genericity and readability of generic markup based languages seem to be the right open answer.
Introducing new abstract interaction metaphors will eventually need to provide users with new content presenters, towards which browsers can switch when needed (e.g. when a new markup ``interactive-video'' is recognized). A simple mechanism for application cooperation can be provided to allow clients to implement such flexible cooperative behaviors. An issue is then to dynamically load on-demand such a cooperative (safe, portable) client from some specific Web location.
An other approach is to offer clients to be dynamically loaded with appropriate (safe, portable) code on-demand, from specific Web locations (using a safe embedded-interpretor, for any existing language). This technique falls in the category of late binding of semantic markups. This problem arises with reconfiguration of ``style sheet'' methods in SGML document presenter applications.
I would like to see such a short research track at the workshop on these topics, as it seems to me an important question for the future Web standard.